Trajectory-based Visual Analytics for Anomalous Human Movement Analysis using Social Media

Abstract

The rapid development and increasing availability of mobile communication and location acquisition technologies allow people to add location data to existing social networks so that people share location-embedded information. For human movement analysis, such location-based social networks have been gaining attention as promising data sources. Researchers have mainly focused on finding daily activity patterns and detecting outliers. However, during crisis events, since the movement patterns are irregular, a new approach is required to analyze the movements. To address these challenges, we propose a trajectory-based visual analytics system for analyzing anomalous human movements during disasters using social media. We extract trajectories from location-based social networks and cluster the trajectories into sets of similar sub-trajectories in order to discover common human movement patterns. We also propose a classification model based on historical data for detecting abnormal movements using human expert interaction.